Computational & Technology Resources
an online resource for computational,
engineering & technology publications 

CivilComp Proceedings
ISSN 17593433 CCP: 97
PROCEEDINGS OF THE SECOND INTERNATIONAL CONFERENCE ON SOFT COMPUTING TECHNOLOGY IN CIVIL, STRUCTURAL AND ENVIRONMENTAL ENGINEERING Edited by: Y. Tsompanakis, B.H.V. Topping
Paper 6
Evolutionary Polynomial Regression as an Alternative Way to Predict the Torsional Strength of Reinforced Concrete Beams A. Fiore^{1}, L. Berardi^{1}, J. Avakian^{2} and G.C. Marano^{2}
^{1}Department of Civil and Environmental Engineering, Technical University of Bari, Bari, Italy
A. Fiore, L. Berardi, J. Avakian, G.C. Marano, "Evolutionary Polynomial Regression as an Alternative Way to Predict the Torsional Strength of Reinforced Concrete Beams", in Y. Tsompanakis, B.H.V. Topping, (Editors), "Proceedings of the Second International Conference on Soft Computing Technology in Civil, Structural and Environmental Engineering", CivilComp Press, Stirlingshire, UK, Paper 6, 2011. doi:10.4203/ccp.97.6
Keywords: reinforced concrete beam, evolutionary polynomial regression, torsional strength, experimental data, building codes, physical insight.
Summary
Determination of ultimate torsion in reinforced concrete (RC) is still an open question because of the number of parameters involved. These parameters are generally characterised by a significant variability, so that only extremely simplified models are used for the description of the physical phenomena. Moreover this results in simplified physical models frequently exhibiting poor agreement with experimental results. Nonetheless, existing models have simple and compact mathematical expressions since they are used by practitioners as building code provisions.
In this paper a different approach for predicting ultimate torsion in RC beams is proposed by using a recent soft computing method, the evolutionary polynomial regression (EPR) technique [1]. EPR can be defined as a nonlinear global stepwise regression, providing symbolic formulæ for models. EPR combines the best features of conventional numerical regression techniques with the effectiveness of genetic programming for constructing symbolic expressions of regression models, and here the input is some experimental results from the technical literature. The procedure output is represented by different formulae to predict the torsional strength of RC beams. The multiobjective search paradigm used by EPR allows the development of a set of formulae showing different complexities of mathematical expressions resulting from different agreement with experimental data. Thus, it is possible to decide a tradeoff between complexity of the mathematical formulae and the accuracy to be used in different design situations. The efficiency of such an approach is tested using the experimental data from sixtyfour rectangular RC beams reported in the technical literature. The input parameters affecting the torsional strength are the crosssectional area of beams, the dimensions of the closed stirrups, the spacing of the stirrups, the crosssectional area of oneleg of a closed stirrup, the yield strength of a stirrup and the longitudinal reinforcement, the steel ratio of the stirrups, the steel ratio of the longitudinal reinforcement and the concrete compressive strength. Final formulations are compared with some building code provisions recently developed for practical structural design [2,3], considering complexity and experimental data fitting. The results show that by using EPR it is possible to obtain models significantly more accurate than the existing building code formulations, although the complexity of their mathematical expressions are comparable. References
purchase the fulltext of this paper (price £20)
go to the previous paper 
